Artificial Intelligence in Selected Domains of Drug Discovery: A Critical Narrative Review

Traditional drug discovery is historically characterized by high attrition rates, escalating financial costs, and decades-long development timelines. As global health challenges-particularly antimicrobial resistance and complex malignancies-intensify, the urgent need for innovative and accelerated therapeutic solutions has never been more critical. Artificial Intelligence (AI) has emerged as a supportive computational framework to address these fundamental bottlenecks, offering advanced computational capabilities to navigate vast chemical spaces and optimize molecular design. While AI-based approaches have demonstrated encouraging performance in specific preclinical settings, their practical impact and limitations require careful, objective evaluation. This critical narrative review examines the application of various artificial intelligence technologies in the design and development of antibiotics, anticancer agents, antibodies, and small-molecule drugs, spanning methodologies from conventional machine learning (ML) to advanced deep learning (DL) models.

A narrative review of studies reporting applications of artificial intelligence in drug discovery and development. It encompassed articles published between 2000 and 2026 and was informed by literature retrieved from multiple electronic databases. The selected studies focused on AI applications in antibiotics, anticancer agents, antibodies, and small-molecule discovery and development. Studies published before 2000, incomplete reports, or those not directly related to pharmaceutical applications of AI were not considered. Review or meta-analysis articles were also excluded from the primary results, though utilized for background context. Although the inclusion criteria covered studies from 2000 to 2026, one earlier study published before 2000 was also included to provide historical context for the early development of neural network applications in molecular biology.

The reviewed literature demonstrates that AI has transitioned from a theoretical concept to a useful framework in early-stage drug discovery, particularly in virtual screening and lead optimization. However, this review identifies a significant "translational gap"; most AI applications remain confined to computational settings, facing challenges in data quality, model interpretability, and a lack of prospective clinical validation. We conclude that while AI significantly accelerates computational efficiency and hypothesis generation, realizing its full potential to combat pressing global health threats requires rigorous experimental integration, standardized data governance, and continuous human expertise to ensure therapeutic efficacy and safety.